Krakatoa Partitioning Multithread?

[romanus]

Hi there,

While sending last job, i got a “bad” surprise. My fastest computer is the slowest while partitioning pflow particles, because it’s only using one cpu. Is this correct or am i doing something wrong while submiting job via deadline?

Cheers,

[Bobo]

If Particle Flow is involved, only one thread will be used. When submitting Partitioning jobs, you have the ability to enable Concurrent Tasks, as long as you have enough memory.
This will cause multiple instances of 3ds Max to be launched on the same machine, each one using more or less one Core. Since neither Max nor Krakatoa require additional licenses to run on the same machine, you can have up to 16 tasks (partitions) processed on a single render node. (but I would suggest starting with 4 and adding more if it seems to help).

Was the actual partitioning time longer than on the other machines? It appears that the 24 core machine is clocked lower than the rest, so it is possible.
But the whole point of Partitioning is to distribute the load to multiple processes and calculate data in parallel, so using the Concurrent Tasks option would be a good idea.
You can enable Concurrent Tasks in the Job Properties of an already submitted job and play with the settings…

[romanus]

@Bobo

If Particle Flow is involved, only one thread will be used. When submitting Partitioning jobs, you have the ability to enable Concurrent Tasks, as long as you have enough memory.
This will cause multiple instances of 3ds Max to be launched on the same machine, each one using more or less one Core. Since neither Max nor Krakatoa require additional licenses to run on the same machine, you can have up to 16 tasks (partitions) processed on a single render node. (but I would suggest starting with 4 and adding more if it seems to help).

Was the actual partitioning time longer than on the other machines? It appears that the 24 core machine is clocked lower than the rest, so it is possible.
But the whole point of Partitioning is to distribute the load to multiple processes and calculate data in parallel, so using the Concurrent Tasks option would be a good idea.
You can enable Concurrent Tasks in the Job Properties of an already submitted job and play with the settings…

Thanks Master Bobo,

Indeed, the 24 core machine is clocked lower.
In terms of concurrent tasks, gonna give it a try! One quick about this… the limit tasks to slave’s task limit checkerbox auto detects the memory limit usage per machine? Because I have 32Gb Ram on one machine and 16 on the others…

Cheers,

[romanus]

Manual info: “Limit Tasks To Slave’s Task Limit: If checked, a slave will not dequeue more tasks than it is allowed to based on its settings” - Based on what settings? RAM memory? I mean, it’s auto? If it’s auto, then we can always use a big number of tasks and let him limit automatically per slave?

Thanks

[Bobo]

@romanus

Manual info: “Limit Tasks To Slave’s Task Limit: If checked, a slave will not dequeue more tasks than it is allowed to based on its settings” - Based on what settings? RAM memory? I mean, it’s auto? If it’s auto, then we can always use a big number of tasks and let him limit automatically per slave?

Thanks

Each Slave has a property called “Concurrent Task Limit Override”. It defaults to 0, in which case the number of Cores is used, unless it is higher than 16, in which case the hard-coded maximum of 16 will be used.
Switch to Super User mode, select a Slave, right-click and select “Modify Slave Properties” (or Ctrl+P) and you will find it.

There is nothing dealing with memory usage, because there is no way to know how much memory a task is actually going to use.
That being said, Krakatoa itself does not use any memory when saving PRTs (or Partitioning). The memory is used only by Max and PFlow.

Hope this helps.